37
$\begingroup$

A search for "lindelof" and a search for "lindelöf" produce different results. In some sense the results are correct, since the first search finds only those posts that mention "lindelof" and the second one only those that mention "lindelöf".

This seems like unfortunate behavior, since many people misspell "lindelöf" as "lindelof". Someone searching for a particular question about lindelöf spaces may not realize that they need to try two separate searches to see all possible results.

It would be nice if the search produced both sets of results for each query.

$\endgroup$
  • 4
    $\begingroup$ There is little that can be done about the difference between "l'Hospital" and "l'Hôpital", but one might hope that searches for "l'Hopital" would find mentions of "l'Hôpital", and vice-versa. At present it does not. $\endgroup$ – MJD Mar 6 '13 at 22:07
  • 1
    $\begingroup$ meta.stackexchange.com/questions/133336 (I upvoted this meta.SO request) $\endgroup$ – user53153 Mar 6 '13 at 22:51
  • $\begingroup$ Google search does not come to our aid on this occasion. Searching for hopital (with the site: parameter as usual) matches on both hopital and hôpital, while searching for hôpital matches only on hôpital. $\endgroup$ – Peter Phipps Mar 7 '13 at 0:12
  • 5
    $\begingroup$ Even more important is Godel and Goedel and Gödel. $\endgroup$ – Asaf Karagila Mar 7 '13 at 1:04
  • 6
    $\begingroup$ In many languages, ö is not an "accented o", but a completely separate letter. Swedish is one such language, and Lindelöf was a Swedish speaking Finn. We who have "strange letters" in our names, myself included, really prefer to have them correctly spelled. $\endgroup$ – mrf Mar 7 '13 at 7:40
  • 9
    $\begingroup$ @mrf I understand that, but the situation we are in is that we have many articles that mention "lindelöf" and many that misspell it as "lindelof". I am not up to the task of correcting all the misspellings, although I try to correct the ones I see. (1 2) If you have a plan for ensuring that Lindelöf's name is always correctly spelled, I would be delighted to hear it. Until then, I think we must be practical about the situation we have. $\endgroup$ – MJD Mar 7 '13 at 7:44
  • 1
    $\begingroup$ Yes, but is this really different from other common misspellings? Searching for "funtion" gives 81 hits (more than the "Lindelof" query). Should these also show up when searching for "function"? $\endgroup$ – mrf Mar 7 '13 at 7:51
  • $\begingroup$ I agree that might also be helpful. But from a purely technical standpoint it seems significantly harder, and perhaps of less benefit. $\endgroup$ – MJD Mar 7 '13 at 7:56
  • 2
    $\begingroup$ Oh dear, there is a risk that I will go on an edit spree to ensure that the name of a compatriot is spelled correctly. Please tap my shoulder, if I go overboard! $\endgroup$ – Jyrki Lahtonen Mar 7 '13 at 12:45
  • $\begingroup$ And "Lindeloef" yields still a different set of results. In one reasonable sense "Lindeloef" is the same as "Lindelöf" but "Lindelof" is different. $\endgroup$ – Michael Hardy Aug 8 '13 at 2:07
  • 1
    $\begingroup$ @MJD Its not much harder. Ignore accents and case then look at everything with a Hamming Distance of 0 or 1 (or possibly 2, but that may be pushing it). It would take a freshman CS Student less than a day to complete such a project. Testing would need to be done to see if these tweeks would be an improvement. $\endgroup$ – Gamma Function Aug 8 '13 at 6:34
  • 9
    $\begingroup$ @MichaelHardy If Lindelöf had been German, Lindeloef would have been an acceptable, but not preferred spelling. However, he was not, and "oe" is not an accepted variant spelling of "ö" in Swedish. There are many other languagues as well with essential diacritical marks. In the age of Unicode, there is no excuse to misspell people's names. $\endgroup$ – mrf Aug 8 '13 at 8:35
  • 2
    $\begingroup$ @Jacob, if you look at everything with Hamming distance 0 or 1, you'll find Weyl when you're searching for Weil, won't you? Yang and Lang? Gilbert and Hilbert? $\endgroup$ – Gerry Myerson Aug 8 '13 at 13:08
  • 4
    $\begingroup$ I wholeheartedly recommend the conversion of some unneeded key (like one of the Win keys on a standard keyboard) into a compose key. Intuitive typing of all kinds of accented letters, no tedious memorizing of code numbers. Examples: [ComposeKey]+["]+[o] produces 'ö', [ComposeKey]+[']+[e] produces 'é', [ComposeKey]+[s]+[s] produces 'ß'. See the Wikipedia article: en.wikipedia.org/wiki/Compose_key/ $\endgroup$ – azimut Sep 27 '13 at 7:47
  • 1
    $\begingroup$ @azimut: My keyboard (for a laptop running Win7) has those keys. An accent followed by a letter gives an accented letter. And the layout being Finnish/Swedish it has separate keys for entering åäö. A consequence of this is that {}[]|\@$ have all been moved behind AltGr. Makes keying in Mathematica (and TeX)-code less fun than it might be with a different keyboard. Win some - lose some. $\endgroup$ – Jyrki Lahtonen Sep 27 '13 at 9:47
5
$\begingroup$

Due to internationalisation efforts, this issue is largely — but not completely — resolved. From the list in the other answer, it appears that most that were accented/unaccented pairs return the same results. For example, "Erdős", "Erdös", and "Erdos" each give the same results.

There are some stragglers, however. While both "Gödel" and "Godel" return the same results, "Goedel" still returns a disjoint set. Similarly, "l'Hôpital"/"l'Hopital" vs. "l'Hospital".

$\endgroup$
  • 1
    $\begingroup$ But the problem is not only with the MSE search engine, but, more globally, with any (browser) search engine that one might use. That's why I often use multiple spellings in the same post (which, alas, some users try to "correct"). $\endgroup$ – Bill Dubuque Jul 17 '15 at 12:37
  • 3
    $\begingroup$ @BillDubuque The question posed was about the Stack Exchange search. If you want to bring up further SEO issues. perhaps start a new question, and not provide off-topic commentary. $\endgroup$ – user642796 Jul 17 '15 at 12:54
  • 1
    $\begingroup$ The question is (implicitly) local but the issue is global. My comment is most certainly not "off topic". $\endgroup$ – Bill Dubuque Jul 17 '15 at 13:10
  • $\begingroup$ Note that even if the set of results are the same, the order when sorting by "relevance" (the default order) is different. For example, searching for "Gödel" returns at the top posts containing the name spelled as "Gödel", while search for "Godel" returns posts containing the incorrect spelling at the top. $\endgroup$ – Najib Idrissi Jul 17 '15 at 13:21
19
$\begingroup$

This does not address the question at all, but perhaps we could make a list of names which are frequently mis-spelled (or have variant spellings/Romanisations).

Please add more to this post, but try to keep the list in "alphabetical" order and include a link to a math.SE search for each variant. Perhaps put in bold the correct/preferred spelling, if any. Also, undeniably incorrect "variants" should perhaps be striked.

Disagreements about preferred spellings should be handled in the comments below and not edit wars.


$\endgroup$
  • 12
    $\begingroup$ Some of these are definitely not stricly misspellings. With Russian names and names from other languages with non-Roman alphabets, there is often no single 'correct' way to write them in Roman letters. Also, at least for German names, if one does not have German characters available (e.g. in website addresses), then it is acceptable to write a vowel followed by an e to represent the vowel with umlaut, and ss for ß. $\endgroup$ – Tara B Mar 7 '13 at 16:19
  • 2
    $\begingroup$ @TaraB True. As Arthur writes: "(or have variant spellings/Romanisations)". Maybe the spellings that are actually wrong could be set in italic. // I briefly considered adding Чебышёв, Лузин, Суслин, Тихонов to the list... $\endgroup$ – user53153 Mar 7 '13 at 16:22
  • 1
    $\begingroup$ @RahulNarain I guess the reasonable criterion is whether those spellings actually exist on Math.SE. (The inclusion of search links supports this). E.g., I did not include Tchebycheff, which is a common French spelling, but does not show up here. $\endgroup$ – user53153 Mar 7 '13 at 17:07
  • 2
    $\begingroup$ @Tara: While several of the variants in the post are not errors (and I admitted as much, as pointed out by 5pm), I feel that it would be beneficial to adhere to some preferred variant of each name. While not everyone will have easy access to the necessary characters at all times, there will always be someone available to put names into their preferred forms via edits. $\endgroup$ – user642796 Mar 8 '13 at 20:01
  • 1
    $\begingroup$ Could we mark those variant spellings which are acceptable with italics or cross out the real mistakes with <s>...</s>? For example, it is unclear to me why Gauß and Gauss are currently marked as equally acceptable, while the list suggests that l'Hôpital should be preferred over l'Hospital, Leibniz over Leibnitz, Mac Lane over MacLane for whatever reasons. $\endgroup$ – Martin Mar 10 '13 at 15:54
  • $\begingroup$ MacLane is correct in some contexts; for example, "Eilenberg–MacLane space". (The name change came later.) $\endgroup$ – Zhen Lin Mar 10 '13 at 21:12
  • $\begingroup$ I have taken @Martin's suggestion and striked some of the variants that (I am fairly certain) are just plain incorrect. I do not claim to have an specialised knowledge about this, so I'll leave it to others to strike variants they know are wrong. (And also to correct any incorrect strikes I may have made.) $\endgroup$ – user642796 Mar 10 '13 at 21:24
  • $\begingroup$ @ZhenLin: Perhaps you could add a note about this for the Mac Lane entry, including any other contexts in which "MacLane" is conventionally used. $\endgroup$ – user642796 Mar 10 '13 at 21:52
  • 5
    $\begingroup$ @Rahul: Not Lindeloef, unless someone can demonstrate that a Finn with a Swedish surname would have used this spelling convention. The German ö is an o with a diacritic; the Swedish ö is a distinct letter, with its own place in the alphabet. $\endgroup$ – Brian M. Scott Mar 11 '13 at 1:57
  • 2
    $\begingroup$ I notice that you’ve not struck any of the just-drop-the-diacritics spellings; is that on grounds of practicality? Because while Erdös is certainly wrong $-$ the vowel is long $-$ it isn’t as wrong as Erdos. $\endgroup$ – Brian M. Scott Mar 11 '13 at 2:02
  • 2
    $\begingroup$ I'm tempted to add both "Lorentz; <s>Lorenz</s>" and "Lorenz; <s>Lorentz</s>"... $\endgroup$ – Henning Makholm Apr 2 '13 at 21:34
  • 2
    $\begingroup$ @ArthurFischer: There are lots of keyboard layouts that provide an easy way to put two dots over a vowel, but no way (short of entering numeric codepoints) to enter an o-with-two-acutes. $\endgroup$ – Henning Makholm Apr 2 '13 at 21:37
  • 1
    $\begingroup$ Any German ö could legitimately be changed to oe, but not to o (ditto ü and ä), so you could strike out those spellings. (I shan't myself, as I don't know off hand which names are German.) $\endgroup$ – TRiG Aug 7 '13 at 13:38
  • 4
    $\begingroup$ I would like to strike out all just-drop-the-diacritics variants of German names (I've already moved them to the end of the list). There highly preferred variant is to use the substitutions 'ä' -> 'ae', 'ö' -> 'oe', 'ü' -> 'ue'. For example, 'König' and 'Koenig' will lead to the same pronunciation by a native German. That's definitely not true for 'Konig' (which additionally, looks quite strange). If you agree, please upvote this comment. If not, comment below. $\endgroup$ – azimut Sep 29 '13 at 12:26
  • 2
    $\begingroup$ @azimut Not really. But if you absolutely want to, dropping the dots is better than replacing ö with oe, ä with ae and å with aa. Just dropping the dots is usually the way people handle Swedish URL:s, and in the spirit of "internationalization", several Swedish companies dropped some diacritical marks from their names. For example "Skånska cementgjuteriet" became just "Skanska". $\endgroup$ – mrf Oct 14 '13 at 19:34
2
$\begingroup$

When I do those two searches lindelöf and lindelof I get the same 720 matches. Only when they are "sorted by relevance" do the lists appear different, since the matches with the exact spelling precede the ones with the approximate spelling. Other ways or sorting more obviously show that the lists are the same.

For more coinfirmation, try searching lindélof and łindelof

(After posting, I note that the question is from 2013, somehow resurrected to the front page.) Is this a change in the search algorithm since 2013?

$\endgroup$
  • $\begingroup$ This had already been noted. tl;dr Yes, now the search considered accented letters equal to non-accented letters, but sorting by relevance prefers exact spelling. $\endgroup$ – Najib Idrissi Jul 28 '16 at 14:16

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .